Dropping an RDFGraph when there is a non-RDFGraph rel table leaves the rel table "dangling" (hashtag2752)

[vikramsubramanian]

Consider an RDFGraph UniKG and another node table Students and a rel table SameAs from Students to UniKG_r. You can drop UniKG even if that would leave Students around. Normally this would give a Binder exception that says: "Cannot delete node table nodeTableX referenced by rel table relTableY."

To replicate:

kuzu> CREATE RDF Graph UniKG;
kuzu> CREATE NODE TABLE Students(name STRING, primary key (name));
kuzu> CREATE REL TABLE SameStudent(FROM Students TO UniKG_r);
kuzu> DROP TABLE UniKG;
kuzu>  MATCH (a:Students)-[s:SameStudent]->(b) RETURN *;
Error: unordered_map::at: key not found

Normally, this would happen:

kuzu> CREATE NODE TABLE nodeTable1 (name STRING, primary key (name));
kuzu> CREATE NODE TABLE nodeTable2 (name STRING, primary key (name));
kuzu> CREATE REL TABLE relTable (FROM nodeTable1 TO nodeTable2);
kuzu> DROP TABLE nodeTable2;
Error: Binder exception: Cannot delete node table nodeTable2 referenced by rel table relTable.

So the system does not allow dropping nodeTable2. )

[mayil-ai[bot]]

Summary: Dropping an RDFGraph when there is a non-RDFGraph rel table leaves the rel table "dangling" causing an error.

Possible Solution

Based on the provided information and code snippets, the issue seems to be related to the deletion of an RDFGraph which leaves a non-RDFGraph relationship table (rel table) dangling, causing an error during a MATCH query after the RDFGraph is dropped. The expected behavior is to prevent the drop with a Binder exception if there are dependencies.

To resolve the issue, the following steps should be taken:

• Modify the dropTableSchema function in src/catalog/catalog_content.cpp to include a check for dependencies before dropping a table schema. If the table is an RDFGraph, ensure that no rel tables are referencing it.
• Implement a check within the dropTableSchema function to throw a BinderException if an attempt is made to drop an RDFGraph that is still referenced by any rel tables.
• Update the DROP TABLE logic in src/binder/bind/bind_ddl.cpp to use the new dependency check logic from dropTableSchema before proceeding with the drop operation.
• Add unit tests in test/ddl/e2e_ddl_test.cpp to verify that attempting to drop an RDFGraph with existing rel table dependencies correctly throws a BinderException with the message "Cannot delete node table nodeTableX referenced by rel table relTableY."
• Ensure that the error message for "unordered_map::at: key not found" is properly handled and that it provides enough context for debugging, such as the missing key and the operation that was being performed.

By implementing these changes, the software should correctly handle the deletion of RDFGraphs and maintain the integrity of the relationships within the database.

Code snippets to check

• src/catalog/catalog_content.cpp

  This file contains the logic for adding and dropping table schemas, which is directly related to the issue of dropping an RDFGraph and leaving a rel table dangling.

• src/binder/bind/ddl/bind_create_rdf_graph.cpp

  This file includes the binder logic for creating RDF graphs, which may need to be reviewed to ensure proper handling of dependencies between RDFGraphs and rel tables.

• src/parser/transform/transform_ddl.cpp

  This file contains the transformation logic for DDL statements, including the creation and dropping of tables, which is relevant to the issue.

• src/binder/bind/bind_graph_pattern.cpp

  This file contains binding logic for graph patterns, which may be relevant when considering how rel tables are bound to RDFGraphs.